File size: 7,932 Bytes
f5fbd23 9d6df4b f5fbd23 9d6df4b 5b475af 9d6df4b 5b475af 9d6df4b f5fbd23 5b475af f5fbd23 5b475af f5fbd23 5b475af f5fbd23 5b475af f5fbd23 9d6df4b f5fbd23 9d6df4b f5fbd23 9d6df4b f5fbd23 9d6df4b f5fbd23 9d6df4b f5fbd23 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 |
import cv2
import numpy as np
from transformers import CLIPProcessor, CLIPModel
import torch
from PIL import Image
import faiss
import pickle
from typing import List, Dict, Tuple
import logging
import gradio as gr
import tempfile
import os
import shutil
class VideoRAGTool:
def __init__(self, model_name: str = "openai/clip-vit-base-patch32"):
"""
Initialize the Video RAG Tool with CLIP model for frame analysis.
"""
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.model = CLIPModel.from_pretrained(model_name).to(self.device)
self.processor = CLIPProcessor.from_pretrained(model_name)
self.frame_index = None
self.frame_data = []
self.logger = self._setup_logger()
def _setup_logger(self) -> logging.Logger:
logger = logging.getLogger('VideoRAGTool')
logger.setLevel(logging.INFO)
handler = logging.StreamHandler()
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
handler.setFormatter(formatter)
logger.addHandler(handler)
return logger
def process_video(self, video_path: str, frame_interval: int = 30) -> None:
"""Process video file and extract features from frames."""
self.logger.info(f"Processing video: {video_path}")
cap = cv2.VideoCapture(video_path)
frame_count = 0
features_list = []
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
if frame_count % frame_interval == 0:
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
image = Image.fromarray(frame_rgb)
inputs = self.processor(images=image, return_tensors="pt").to(self.device)
image_features = self.model.get_image_features(**inputs)
self.frame_data.append({
'frame_number': frame_count,
'timestamp': frame_count / cap.get(cv2.CAP_PROP_FPS)
})
features_list.append(image_features.cpu().detach().numpy())
frame_count += 1
cap.release()
if not features_list:
raise ValueError("No frames were processed from the video")
features_array = np.vstack(features_list)
self.frame_index = faiss.IndexFlatL2(features_array.shape[1])
self.frame_index.add(features_array)
self.logger.info(f"Processed {len(self.frame_data)} frames from video")
def query_video(self, query_text: str, k: int = 5) -> List[Dict]:
"""Query the video using natural language and return relevant frames."""
self.logger.info(f"Processing query: {query_text}")
inputs = self.processor(text=[query_text], return_tensors="pt").to(self.device)
text_features = self.model.get_text_features(**inputs)
distances, indices = self.frame_index.search(
text_features.cpu().detach().numpy(),
k
)
results = []
for i, (distance, idx) in enumerate(zip(distances[0], indices[0])):
frame_info = self.frame_data[idx].copy()
frame_info['relevance_score'] = float(1 / (1 + distance))
results.append(frame_info)
return results
class VideoRAGApp:
def __init__(self):
self.rag_tool = VideoRAGTool()
self.current_video_path = None
self.processed = False
self.temp_dir = tempfile.mkdtemp()
def __del__(self):
"""Cleanup temporary files on deletion"""
if hasattr(self, 'temp_dir') and os.path.exists(self.temp_dir):
shutil.rmtree(self.temp_dir, ignore_errors=True)
def process_video(self, video_file):
"""Process uploaded video and return status message"""
try:
if video_file is None:
return "Please upload a video first."
# video_file is now a file path provided by Gradio
video_path = video_file.name
# Create a copy in our temp directory
temp_video_path = os.path.join(self.temp_dir, "current_video.mp4")
shutil.copy2(video_path, temp_video_path)
self.current_video_path = temp_video_path
self.rag_tool.process_video(self.current_video_path)
self.processed = True
return "Video processed successfully! You can now ask questions about the video."
except Exception as e:
self.processed = False
return f"Error processing video: {str(e)}"
def query_video(self, query_text):
"""Query the video and return relevant frames with descriptions"""
if not self.processed:
return None, "Please process a video first."
try:
results = self.rag_tool.query_video(query_text, k=4)
frames = []
captions = []
cap = cv2.VideoCapture(self.current_video_path)
for result in results:
frame_number = result['frame_number']
cap.set(cv2.CAP_PROP_POS_FRAMES, frame_number)
ret, frame = cap.read()
if ret:
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frames.append(Image.fromarray(frame_rgb))
caption = f"Timestamp: {result['timestamp']:.2f}s\n"
caption += f"Relevance: {result['relevance_score']:.2f}"
captions.append(caption)
cap.release()
return frames, "\n\n".join(captions)
except Exception as e:
return None, f"Error querying video: {str(e)}"
def create_interface(self):
"""Create and return Gradio interface"""
with gr.Blocks(title="Video Chat RAG") as interface:
gr.Markdown("# Video Chat RAG")
gr.Markdown("Upload a video and ask questions about its content!")
with gr.Row():
video_input = gr.File(
label="Upload Video",
file_types=["video"],
)
process_button = gr.Button("Process Video")
status_output = gr.Textbox(
label="Status",
interactive=False
)
with gr.Row():
query_input = gr.Textbox(
label="Ask about the video",
placeholder="What's happening in the video?"
)
query_button = gr.Button("Search")
with gr.Row():
gallery = gr.Gallery(
label="Retrieved Frames",
show_label=True,
elem_id="gallery",
columns=[2],
rows=[2],
height="auto"
)
captions = gr.Textbox(
label="Frame Details",
interactive=False
)
process_button.click(
fn=self.process_video,
inputs=[video_input],
outputs=[status_output]
)
query_button.click(
fn=self.query_video,
inputs=[query_input],
outputs=[gallery, captions]
)
return interface
# Initialize and create the interface
app = VideoRAGApp()
interface = app.create_interface()
# Launch the app
if __name__ == "__main__":
interface.launch() |